TY - GEN
T1 - Automatic Crack Detection Approach for the Offshore Flare System Inspection
AU - Tosawadi, Teepakorn
AU - Choppradit, Pakcheera
AU - Sookpong, Satida
AU - Phimsiri, Sasin
AU - Suttichaya, Vasin
AU - Utintu, Chaitat
AU - Thamwiwatthana, Ek
N1 - Publisher Copyright:
© 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.
PY - 2023
Y1 - 2023
N2 - Asset inspection using unmanned aerial vehicle (UAV) is gaining more attention in surveillance and exploratory engineering. UAV surveillance enables higher efficiency in terms of operating budget and time. This technique is usually used to gather asset information for detecting the anomalies, such as cracks and rust. However, cracking detection on the images from UAV’s camera also requires expertise to examine each image, which is time-consuming. To mitigate the issue, automatic crack detection on the flare stack images is proposed in this paper. This research used UAV with high-resolution cameras to collect the flare stack images from the oil refineries. The proposed method introduced crack detection using three object detection models, YOLOv5, YOLOv7, and DINO. Additionally, the Sobel operator and Canny edge detector are also applied to preprocessing images. With an inference time of 50 ms and performance of 65.4% mAP0.5 and 32.2% mAP0.5:0.95, the results show that YOLOv7-W6 on RGB images had the best performance and fastest inference time.
AB - Asset inspection using unmanned aerial vehicle (UAV) is gaining more attention in surveillance and exploratory engineering. UAV surveillance enables higher efficiency in terms of operating budget and time. This technique is usually used to gather asset information for detecting the anomalies, such as cracks and rust. However, cracking detection on the images from UAV’s camera also requires expertise to examine each image, which is time-consuming. To mitigate the issue, automatic crack detection on the flare stack images is proposed in this paper. This research used UAV with high-resolution cameras to collect the flare stack images from the oil refineries. The proposed method introduced crack detection using three object detection models, YOLOv5, YOLOv7, and DINO. Additionally, the Sobel operator and Canny edge detector are also applied to preprocessing images. With an inference time of 50 ms and performance of 65.4% mAP0.5 and 32.2% mAP0.5:0.95, the results show that YOLOv7-W6 on RGB images had the best performance and fastest inference time.
KW - Crack detection
KW - Flare stack crack
KW - Flare stack inspection
UR - https://www.scopus.com/pages/publications/85180745675
U2 - 10.1007/978-3-031-50158-6_31
DO - 10.1007/978-3-031-50158-6_31
M3 - Conference contribution
AN - SCOPUS:85180745675
SN - 9783031501579
T3 - Lecture Notes in Networks and Systems
SP - 314
EP - 323
BT - Intelligent Computing and Optimization - Proceedings of the 6th International Conference on Intelligent Computing and Optimization 2023 ICO2023
A2 - Vasant, Pandian
A2 - Shamsul Arefin, Mohammad
A2 - Panchenko, Vladimir
A2 - Thomas, J. Joshua
A2 - Munapo, Elias
A2 - Weber, Gerhard-Wilhelm
A2 - Rodriguez-Aguilar, Roman
PB - Springer Science and Business Media Deutschland GmbH
T2 - 6th International Conference on Intelligent Computing and Optimization, ICO 2023
Y2 - 27 April 2023 through 28 April 2023
ER -